Vous êtes sur la page 1sur 31

ARTICLE IN PRESS

Journal of Accounting and Economics 44 (2007) 193–223


www.elsevier.com/locate/jae

Determinants of weaknesses in internal control


over financial reporting$
Jeffrey Doylea, Weili Geb, Sarah McVayc,
a
College of Business, Utah State University, 3500 Old Main Hill, Logan, UT 84322, USA
b
University of Washington Business School, University of Washington, Mackenzie Hall, Box 353200,
Seattle, WA 98195, USA
c
Stern School of Business, New York University, 44 West Fourth Street, Suite 10– 94, New York, NY 10012, USA

Received 2 March 2005; received in revised form 15 May 2006; accepted 26 October 2006
Available online 26 December 2006

Abstract

We examine determinants of weaknesses in internal control for 779 firms disclosing material
weaknesses from August 2002 to 2005. We find that these firms tend to be smaller, younger,
financially weaker, more complex, growing rapidly, or undergoing restructuring. Firms with more
serious entity-wide control problems are smaller, younger and weaker financially, while firms with
less severe, account-specific problems are healthy financially but have complex, diversified, and
rapidly changing operations. Finally, we find that the determinants also vary based on the specific
reason for the material weakness, consistent with each firm facing their own unique set of internal
control challenges.
r 2006 Elsevier B.V. All rights reserved.

JEL Classification: M41

Keywords: Internal control; Material weakness; Sarbanes-Oxley

$
We would like to thank Eli Bartov, Donal Byard, Patty Dechow, Ilia Dichev, Mei Feng, Nader Hafzalla,
Gene Imhoff, Kalin Kolev, Andy Leone (the discussant and reviewer), Feng Li, Russ Lundholm, Suzanne
Morsfield, Kyle Peterson, Stephen Ryan, Cathy Shakespeare, and Jerry Zimmerman (the editor) for their helpful
comments and suggestions. This paper has also benefited from comments by workshop participants at the 2005
4-School Conference at Columbia University, the 2005 AAA Midwest Regional Meeting, the 2005 AAA Annual
Meeting, and the University of Michigan. Professor Doyle acknowledges financial assistance from the David
Eccles School of Business at the University of Utah. All errors are our own.
Corresponding author. Tel.: +1 212 998 0040; fax: +1 212 995 4004.
E-mail address: smcvay@stern.nyu.edu (S. McVay).

0165-4101/$ - see front matter r 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.jacceco.2006.10.003
ARTICLE IN PRESS
194 J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223

1. Introduction

In this paper, we examine the determinants of material weaknesses in internal control


over financial reporting. A material weakness in internal control is defined as ‘‘a significant
deficiency, or combination of significant deficiencies, that results in more than a remote
likelihood that a material misstatement of the annual or interim financial statements will
not be prevented or detected’’ (PCAOB, 2004).1 We use a sample of companies that
disclosed material weaknesses in internal control over financial reporting under Sections
302 and 404 of the Sarbanes-Oxley Act of 2002 from August 2002 to 2005.2 Under Section
302, SEC registrants’ executives are required to certify that they have evaluated the
effectiveness of their internal controls over financial reporting. If management identifies a
material weakness in their controls, they are precluded from reporting that the controls are
effective and must disclose the identified material weakness (SEC, 2002, 2004). Section 404
requires that each annual report include an assessment by management of the effectiveness
of the internal control structure and procedures of the issuer for financial reporting that is
attested to by the firm’s public accountants.
Although firms were required to maintain an adequate system of internal control before
the enactment of Sarbanes-Oxley, they were only required to publicly disclose deficiencies
if there was a change in auditor (SEC, 1988). While prior research studies this limited set of
disclosures (Krishnan, 2005), there is little evidence regarding internal control quality for
firms in general under the new Sarbanes-Oxley regime.
We investigate whether material weaknesses in internal control are associated with
(1) firm size, measured by market value of equity; (2) firm age, measured by the number of
years the firm has CRSP data; (3) financial health, measured by an aggregate loss indicator
variable and a proxy for the likelihood of bankruptcy based on the hazard model
developed by Shumway (2001); (4) financial reporting complexity, measured by the
number of special purpose entities reported, the number of segments reported, and the
existence of a foreign currency translation; (5) rapid growth, measured by merger
and acquisition expenditures and extreme sales growth; (6) restructuring charges; and
(7) corporate governance, measured using the governance score developed by Brown and
Caylor (2006).

1
A significant deficiency is defined as ‘‘a control deficiency, or combination of control deficiencies, that
adversely affects the company’s ability to initiate, authorize, record, process, or report external financial data
reliably in accordance with generally accepted accounting principles such that there is more than a remote
likelihood that a misstatement of the company’s annual or interim financial statements that is more than
inconsequential will not be prevented or detected’’ (PCAOB, 2004, Auditing Standard 2, Paragraph 9). A
‘‘significant deficiency’’ and a ‘‘material weakness’’ are both deficiencies in the design or operation of internal
controls, but significant deficiencies are less severe and are not required to be publicly disclosed under Sections 302
or 404 (SEC, 2004).
2
Section 404 became effective for fiscal years ending after November 15, 2004 for accelerated filers, which
generally includes public firms with a market capitalization of at least $75 million (the due date was extended an
additional 45 days for accelerated filers with a market capitalization of less than $700 million in November 2004).
For non-accelerated filers, Section 404 will be effective for years ending after December 15, 2007 and auditor
attestation will be required for years ending after December 15, 2008. Non-accelerated filers, however, must still
evaluate their controls and disclose any material weaknesses under Section 302. To the extent that our inclusion of
Section 404 disclosures biases our sample toward larger firms, the inclusion of market value of equity in our
multivariate analyses should act as a control. In untabulated results we also replicate our results using only the
Section 302 disclosures and find qualitatively similar results.
ARTICLE IN PRESS
J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223 195

Our sample is comprised of 970 unique firms that reported at least one material
weakness from August 2002 to 2005, of which 779 have Compustat data. We identify these
firms through a combination of a search of Compliance Week, a website which tracks
internal control disclosures after Sarbanes-Oxley, and a search of 10-K filings in the
EDGAR database.
For the full sample, we find that material weaknesses in internal control are more likely
for firms that are smaller, younger, financially weaker, more complex, growing rapidly,
and/or undergoing restructuring. These firm-specific characteristics seem to create
challenges for companies in maintaining a strong system of internal controls. Our findings
also appear to be economically significant in identifying firms with material weaknesses.
For example, the joint marginal effect of our main model (i.e., the change in the predicted
probability of a material weakness when altering the independent variables in the predicted
direction between the 1st and 3rd quartiles or between zero and one for indicator variables)
greatly increases the predicted probability of a material weakness—from 3.75 percent to
26.41 percent.
In this paper, we focus solely on material weaknesses for two reasons. First, it is the
most severe type of internal control deficiency, and thus offers the greatest power for our
determinants tests. Second, the disclosure of material weaknesses is effectively mandatory,
while the disclosure of ‘‘significant deficiencies’’ is unambiguously voluntary.3 Focusing on
these more mandatory disclosures helps avoid self-selection issues associated with
voluntary disclosures. Although disclosures of material weaknesses are effectively
mandatory, it is possible that individual firms or auditors apply different materiality
standards in deciding what to disclose. While we do not have a model of the materiality
threshold of material weaknesses (Mayper, 1982; Mayper et al., 1989; Messier et al., 2005),
our determinants results are similar to those documented by Ashbaugh-Skaife et al. (2007)
who examine all types of significant deficiencies (i.e., not just those internal control
weaknesses that meet the threshold to be classified as ‘‘material weaknesses’’) and find that
firms disclosing significant deficiencies typically have more complex operations, recent
changes in organization structure, more accounting risk exposure, and fewer resources to
invest in internal control. Therefore, it appears that our results extend to a broader sample
that does not rely on a potentially subjective judgment of what constitutes a ‘‘material
weakness,’’ although it is still possible that the broader sample in Ashbaugh-Skaife et al.
(2007) suffers from the same concern.4 Since Ashbaugh-Skaife et al. (2007) focus on all
significant deficiencies, including unambiguously voluntary disclosures, they also include

3
Although disclosure of material weaknesses is definitely mandatory under Section 404 (SEC, 2003), there is
some ambiguity regarding whether Section 302 certifications require public disclosure of material weaknesses. For
example, Question 9 of the SEC’s Frequently Asked Questions (SEC, 2004) seems to imply that firms should only
‘‘carefully consider’’ whether to publicly disclose material weaknesses. However in Question 11 they state without
reserve that ‘‘A registrant is obligated to identify and publicly disclose all material weaknesses.’’ Confusion arises
due to the existence of two largely overlapping definitions of controls (‘‘disclosure controls and procedures’’ and
‘‘internal controls over financial reporting’’), two reporting regimes (Sections 302 and 404), and two tiers of
reporting requirements (accelerated vs. non-accelerated filers). Although it is possible that some firms might
interpret the material weakness disclosure requirement under Section 302 as voluntary, our reading of the bulk of
SEC guidance and many firms’ begrudging material weakness disclosures seems to indicate that most firms are
treating the disclosure as mandatory.
4
Moreover, prior research on materiality thresholds for internal control problems finds that the type of problem
(e.g., segregation of duties), rather than firm characteristics, is the best indicator of whether the internal control
problem will be classified as a material weakness (Mayper, 1982, p. 782).
ARTICLE IN PRESS
196 J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223

additional variables to model the choice to disclose in their analyses. Since our focus is on
material weakness disclosures, we do not include these variables in our main analysis. In
untabulated results, our results are robust to their inclusion, though sales growth weakens
considerably in the more restricted sample (with or without the additional variables).5
In addition to our general findings about material weakness firms, discussed above,
which complement and corroborate the findings of concurrent studies, we differ from
Ashbaugh-Skaife et al. (2007) and others by examining the specific types of material
weaknesses disclosed, and how the determinants of internal control problems differ based
on these types. We find that the type of internal control problem is an important factor
when examining determinants, and thus should be considered by future research on
internal control. Specifically, while we focus on material weaknesses, the most severe
internal control problems, these weaknesses vary widely with respect to severity and
underlying reason. For example, consider the two following material weakness disclosures:

As part of the annual audit process, a material weakness was identified in our
controls related to the application of generally accepted accounting principles,
specifically related to the classification of the Company’s short-term investments,
resulting in the Company reclassifying approximately $34 million of cash and cash
equivalents to short-term investmentsy (I-Flow Corporation, 12/31/04 10-K).
The material weaknesses identified by the independent registered accounting firm
include the following weaknesses in certain divisions of the Company: (1) Failure to
reconcile certain general ledger accounts on a timely and regular basis and lack
of management review of certain reconciliations. (2) Inconsistent application
of accounting policies, including capitalization policies and procedures for
determining unrecorded liabilities. (3) Failure of financial management in certain
operating segments to properly supervise personnel, enforce and follow policies and
procedures, and perform their assigned duties. (4) Lack of adequately staffed
accounting departments (Evergreen Holdings, Inc., 2/29/2004 10-K/A).

While I-Flow’s disclosure relates to an account-specific balance sheet classification error,


Evergreen’s disclosure speaks of larger, more pervasive problems in the company. This
distinction is deemed to be important by Moody’s, the bond rating company. Moody’s
posits that while account-specific weaknesses are auditable, company-level weaknesses are
more difficult to audit around and call into question not only management’s ability to
prepare accurate financial reports but also its ability to control the business (Doss and
Jonas, 2004). We investigate whether the determinants of these two types of weaknesses
differ.
We find that firms that report account-specific weaknesses tend to be larger, older, and
financially healthier than firms that report company-level weaknesses. They also have more
complex and diversified business operations and are growing more rapidly. The complexity
of their operating environment, along with the rapid change evidenced by merger and
acquisition activity and high sales growth, seems to hinder these firms in maintaining
5
The additional proxies for the incentives to discover and disclose internal control problems include the size of
the auditor (consistent with Ge and McVay, 2005), the existence of a past restatement (which could also be
evidence of lower accruals quality in the presence of weak internal controls, Doyle et al., 2007), the level of
ownership concentration, and whether or not the firm operates in a litigious industry. The latter two variables
were not significant in our regressions.
ARTICLE IN PRESS
J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223 197

adequate account-specific internal controls. In contrast, firms with company-wide


problems seem to lack the resources or experience to maintain comprehensive control
systems.
We also examine whether the determinants differ based on whether the firm attributes its
material weakness to staffing issues (e.g., segregation of duties), complexity issues (e.g.,
trouble in calculating the deferred tax provision) or more general issues (e.g., lack of
supporting documentation). Not surprisingly, firms disclosing staffing problems are more
likely to be smaller and younger than other firms disclosing material weaknesses. These
firms also tend to be the weakest financially, with the highest incidence of losses and the
highest bankruptcy risk. Resource constraints likely hinder the ability of such firms to
adequately staff their operations with competent personnel.
Firms disclosing material weaknesses related to complexity are the largest and oldest
companies of the three groups and have the most sophisticated and diversified operations.
In addition, when compared to the average Compustat firm, these firms continue to have
more diversified and complex operations, and also tend to be weaker financially and have
higher restructuring charges. Thus, complex operations, combined with relatively poor
financial health and a quickly changing environment, appear to yield difficult financial
reporting issues for these firms.
When examining firms providing more general material weakness disclosures, we find
that each of the constructs examined tends to be associated with these firm disclosures,
consistent with this subgroup containing many differing weaknesses (e.g., inadequate
reconciliation procedures, revenue recognition problems, or a complete lack of policies and
procedures in place). As a final analysis, we examine only those firms with material
weaknesses related to revenue recognition problems and find that these disclosures are
negatively associated with our proxy for good corporate governance.
In Section 2, we discuss the new requirements on internal control disclosures,
prior research, and our hypotheses. In Section 3, we discuss our sample selection
procedure and the data items used as construct proxies. In Section 4, we describe the
methodology used to test our hypotheses and discuss the results. We summarize and
conclude in Section 5.

2. Background of internal control over financial reporting, prior research, and hypotheses

2.1. Background and prior research

Internal control over financial reporting has long been recognized as an important
feature of a company (see Kinney et al., 1990; Kinney, 2000, 2001).6 However, prior to
Sarbanes-Oxley, standards in place were very limited in scope. The sole statutory
regulation of internal control over all SEC registrants was the Foreign Corrupt Practices
6
Internal control over financial reporting ‘‘includes those policies and procedures that: (1) pertain to the
maintenance of records that, in reasonable detail, accurately and fairly reflect the transactions and dispositions of
the assets of the company; (2) provide reasonable assurance that transactions are recorded as necessary to permit
preparation of financial statements in accordance with generally accepted accounting principles, and that receipts
and expenditures of the company are being made only in accordance with authorizations of management and
directors of the company; and (3) provide reasonable assurance regarding prevention or timely detection of
unauthorized acquisition, use or disposition of the company’s assets that could have a material effect on the
financial statements’’ (PCAOB, 2004).
ARTICLE IN PRESS
198 J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223

Act (FCPA) of 1977, and the only required public disclosure of significant internal control
deficiencies for all SEC firms was in the firm’s 8-K, when disclosing a change in auditors
(SEC, 1988; Geiger and Taylor, 2003; Krishnan, 2005). Within the banking industry, the
Federal Deposit Insurance Corporation Improvement Act (FDICIA) of 1991 requires
banks operating in the United States to file an annual report with regulators in which
management attests to the effectiveness of their controls, and their independent public
accountants attest to and separately report on management’s assertions.
Similarly, under Section 404 of Sarbanes-Oxley (effective November 15, 2004 for
accelerated filers) managers must review and provide an annual report on their internal
controls, assessing the effectiveness of the internal control structure and procedures.
However, even before the implementation of Section 404, firms began disclosing material
weaknesses in their controls in response to Section 302 of Sarbanes-Oxley, under which the
company executives are required to certify in the periodic reports (e.g., the 10-Qs and
10-Ks) filed with the SEC that their systems of controls are effective and report any
significant changes in internal control.7
The details of internal control problems are most often provided in Item 9A—Controls
and Procedures in firms’ 10-Ks and Item 4—Controls and Procedures in firms’ 10-Qs. In
addition, managers often discuss internal control problems under Risk Factors in the
MD&A. At least 970 firms have disclosed at least one material weakness in internal control
from August 2002 to August 2005 under both Sections 302 and 404. We provide several
additional examples of material weakness disclosures in Appendix A.
Both Sections 302 and 404 use definitions of ‘‘effective’’ internal control similar to those
developed in 1992 by the Committee of Sponsoring Organizations (COSO) of the
Treadway Commission. The SEC thus defines internal control as ‘‘a process, effected by an
entity’s board of directors, management and other personnel, designed to provide
reasonable assurance regarding the reliability of financial reporting.’’ Although the COSO
framework broadly defines internal control in terms of achieving (1) the effectiveness and
efficiency of operations, (2) reliability of financial reporting, and (3) compliance with
applicable laws and regulations (Statements on Auditing Standards, Section 319),
Sarbanes-Oxley only pertains to internal control related to the reliability of financial
reporting.8
Internal control is a major focus of recent regulatory changes under Sarbanes-Oxley.
However, empirical research on the determinants of internal control quality prior to
Sarbanes-Oxley is extremely limited. The most direct evidence is provided by Krishnan

7
In its final rules, the SEC adopted the commonly known definition for a material weakness under existing
GAAS and attestation standards (Interim Auditing Standards AU 325.15, PCAOB; see also SEC, 2003 and
Krishnan, 2005). Although the material weakness definition was slightly updated by the PCAOB in March 2004 in
its issuance of Auditing Standard 2 (Paragraph 10), it was essentially unchanged from before. Furthermore, about
90% of our sample disclosures are after March 2004, resulting in a fairly standard ‘‘material weakness’’ definition
across time.
8
For Section 302, the SEC also refers to ‘‘disclosure controls and procedures,’’ which largely overlap with
‘‘internal control over financial reporting.’’ The definition encompasses ‘‘the quality and timeliness of disclosure’’
to the SEC and is intended to include ‘‘material non-financial information, as well as financial information’’ (SEC,
2002), but may exclude some internal controls such as ‘‘safeguarding of assets’’ (SEC, 2003). In practice, there
seems to be quite a bit of confusion about how the concepts differ, and it seems that most firms generally refer to
the ‘‘internal control over financial reporting’’ definition in their disclosures. Our results are robust to the
exclusion of the Section 404 disclosures and their potentially different control definition.
ARTICLE IN PRESS
J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223 199

(2005). She examines 128 internal control deficiencies (including significant deficiencies
that are not classified as material weaknesses) reported from 1994 to 2000 in the
8-Ks of firms that changed auditors. Her focus is on the association between audit
committee quality and internal control quality, which she finds to be positively related.
However, her sample is limited to firms that changed auditors. We present a much broader
study of the determinants of internal control problems with our sample of 779 unique
firms disclosing material weaknesses in the three years following the effective date of
Section 302.
To provide evidence on the pervasiveness of material weakness disclosures prior to
Sarbanes-Oxley we search all 10-Ks in the EDGAR database for the three years prior to
Section 302 (from August 1, 1999 to August 1, 2002) using the keywords ‘‘material
weakness’’ and ‘‘material weaknesses.’’ We identify 61 distinct disclosures of material
weaknesses. Of these 61 disclosures, 40 are listed under Item 9—Changes in and
Disagreements with Accountants on Accounting and Financial Disclosure, and pertain to
a change in auditor, the mandatory disclosure requirement discussed above. Among the 21
voluntary disclosures noted, six were disclosed in conjunction with a restatement of the
financial statements, four with the disclosure of theft or fraud, and two were identified
and disclosed by new senior management. Clearly there has been a marked increase in the
disclosure of material weaknesses following the passage of Sarbanes-Oxley, opening the
door to many new studies in this area.9
Prior to Sarbanes-Oxley, many studies opted to provide indirect evidence on internal
control. Kinney and McDaniel (1989) examine characteristics of 73 firms that correct
previously reported quarterly earnings from 1976 to 1985. They posit that a restatement
implies a breach in the firm’s internal control system, and find that both firm size and firm
profitability are negatively associated with these restatements in univariate tests. DeFond
and Jiambalvo (1991) examine 41 firms with prior period adjustments from 1977 to 1988
and use firm size as a proxy for the strength of a firm’s internal controls. While firm size is
weakly negatively associated with prior period adjustments in univariate tests, they find
that firm size is not a statistically significant variable in their multivariate regression
analysis. Finally, McMullen et al. (1996) proxy for weak internal control with both SEC
enforcement actions and corrections of previously reported earnings. Their focus is on
whether weak internal control firms voluntarily report on internal control. They find that
small firms with weak internal control are less likely than other small firms to provide
voluntary reports on internal control.
As noted above, restatements are often viewed as indicative of internal control
problems. However, little research examines the determinants of restatements. Again,
Kinney and McDaniel (1989) and DeFond and Jiambalvo (1991) document negative
univariate associations between restatements and size and profitability. In addition,
Richardson et al. (2003) examine the determinants of income-decreasing restatements
using a sample of 225 restatement firms with 440 restatements from 1971 to 2000. The
authors find that neither size nor profitability varies between their test and control firms.
9
For example, Bryan and Lilien (2005) examine firm characteristics such as firm size and beta, Chan et al. (2005)
examine if firms reporting material weaknesses in internal control under Section 404 have more earnings
management and lower return-earnings associations compared to other firms, Doyle et al. (2007) examine the
accruals quality of material weakness firms, Hogan and Wilkins (2005) examine earnings management and audit
fees, and Beneish et al. (2006) and Ogneva et al. (2006) examine the association between implied cost of equity and
internal control effectiveness.
ARTICLE IN PRESS
200 J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223

Their focus is largely on the incentives to manage earnings (e.g., external financing and
meeting the analyst forecast). As discussed later in Section 4.4, we also perform our
tests on only those material weakness firms that do not contemporaneously restate
their financial statements in order to ensure that we are not merely documenting the
determinants of restatements.

2.2. Hypotheses

As reviewed above, there is limited guidance from prior research regarding the
determinants of internal control quality. Thus, while we attempt to incorporate this
literature in the formulation of our hypotheses, our study may be viewed as exploratory in
nature, and a first step in examining the determinants of internal control quality. To begin,
prior research hypothesizes that firm size may be a determinant of good internal control
(e.g., Kinney and McDaniel, 1989; DeFond and Jiambalvo, 1991), though the evidence is
mixed (DeFond and Jiambalvo, 1991; Krishnan, 2005). Intuitively, large firms likely have
more financial reporting processes and procedures in place and are more likely to have an
adequate number of employees to ensure proper segregation of duties.10 Larger firms are
also more likely to enjoy economies of scale when developing and implementing internal
control systems. Moreover they tend to have greater resources to spend on internal
auditors or consulting fees, which may aid in the generation of strong internal control. For
example, there is a strong positive association between non-audit fees and firm size (e.g.,
DeFond et al., 2002; Frankel et al., 2002). One possible confounding factor in prior
research is that large firms also tend to be more complex and engage in a larger number
and variety of transactions. However, as discussed below, we explicitly control for
complexity in our tests. Thus, we expect to find fewer control weaknesses in larger firms,
after controlling for complexity. We measure firm size (MARKETCAP) as the log of the
firm’s market value of equity.11
Another factor that likely determines the processes and procedures in place is the age of
the firm. The older the firm, the more likely they are to have ‘‘ironed out the kinks’’ in their
internal control procedures. Thus, we expect to find fewer control weaknesses in older
firms. We define FIRM AGE as the log of the number of years the firm has been public,
measured by the number of years the firm has price information on CRSP.
A third determinant of strong internal control is expected to be a firm’s financial health.
Poorly performing firms simply may not be able to adequately invest time and/or money in
proper controls. Good internal control requires both financial resources and management
time, and this may not be a priority for firms that are concerned about simply staying in
business. Consistent with this hypothesis, past research finds that financial reporting errors
10
Weak internal control may occur in equilibrium, especially for small firms. For example, Universal Security
Instruments, Inc., reported a material weakness in their internal control regarding a lack of segregation of duties.
However, ‘‘ymanagement has decided that y the risks associated with such lack of segregation are insignificant
and the potential benefits of adding employees to clearly segregate duties do not justify the expenses associated
with such increases.’’
11
We present the log of the market value of equity as our size proxy since we span the effective date of Section
404 for accelerated filers, a definition largely based on market value. We also use market capitalization because
our underlying construct is firm resources, and we think market capitalization best captures the resources
available to the firm. Size continues to be a negative predictor of material weakness disclosures if we examine the
log of total assets or the log of total book value, but is not significant if we use the log of sales.
ARTICLE IN PRESS
J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223 201

are negatively associated with performance (DeFond and Jiambalvo, 1991) and that the
existence of a loss is positively associated with reporting an internal control problem in
audit-change firms (Krishnan, 2005). We expect to find fewer internal control weaknesses
in firms with stronger financial health. We examine two financial resource measures,
AGGREGATE LOSS (whether or not the sum of earnings before extraordinary items for
years t and t–1 is negative) and BANKRUPTCY RISK, the decile rank of the percentage
probability of bankruptcy from the hazard default prediction model developed by
Shumway (2001).12
While we expect the size, age, and financial resources of the firm to affect its ability to
establish proper internal controls, the need for internal controls is unique to each firm’s
particular operating environment.13 As a firm engages in more complex transactions and
has more diverse operations, we expect the need for internal control to be higher, and thus
expect the complexity of the firm to be a driver of internal control weaknesses. Consider
the complexity introduced by having multiple geographic or business divisions. These
companies face challenges when implementing internal control consistently across different
divisions and when consolidating information for financial statements. For each division,
different factors might affect the implementation of adequate internal control. For
example, for a multinational company, the local institutional and legal environment of
each location might differ, and thus affect the effectiveness of internal control.14 We
examine complexity using three measures: the log of the number of special purpose entities
associated with the firm, SPEs, the log of the sum of the number of operating and
geographic segments, SEGMENTS, and the existence of a foreign currency adjustment,
FOREIGN TRANSACTIONS (e.g., DeFond et al., 2002; Bushman et al., 2004).15
A fifth possible determinant of internal control weaknesses is rapid growth. A quickly
growing firm may outgrow any internal controls it has in place, and may require time to
establish new procedures (Kinney and McDaniel, 1989; Stice, 1991). New personnel,
processes, and technology are usually needed to match the internal control with the
firm’s growth. For example, in their 2003 10-K, MarkWest Energy Partners disclosed
‘‘inadequate implementation of uniform controls over certain acquired entities and

12
Results are similar if we use a three-year earnings aggregation (t, t–1, t–2). We present the two-year
aggregation to minimize data requirements. Shumway’s model uses accounting information, market returns, and
return volatility to estimate the probability of bankruptcy in any given period. Results are stronger if we use the
Altman Z-score instead of the Shumway score. We present the Shumway score as this variable has more available
observations.
13
For example, Glass and Lewis (2004) state ‘‘ythe size of the company, quality of its staff, risk management
processes, the type of information technology systems, complexity of product lines including marketing channels,
geographical dispersion, nature of the business (e.g. manufacturing versus service) as well as many other factors
can affect the types of internal control necessary. Most often, small businesses will not require the type of complex
internal control systems that are required of large international conglomerates engaging in extensive risk
management techniques and financial instruments.’’
14
For example, Baxter International Inc., one of our sample companies, disclosed a material weakness that
occurred in their Brazilian division in their 2003 10-K. For firms with complex organizations, it may not be cost
effective for them to maintain a high level of controls at all locations or divisions.
15
We thank Mei Feng, Jeff Gramlich, and Sanjay Gupta for providing the special purpose entity data. We use
the sum of the number of operating and geographic segments for two reasons. First, in most cases, companies
report either operating segments or geographic segments; second, in some cases, companies report location related
segments (e.g., Japan) as their operating segments. Thus the distinction between operating and geographic
segments is not very clear. However, the results are similar if we use the number of operating and geographic
segments as two separate variables.
ARTICLE IN PRESS
202 J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223

operations’’ as one aspect of their material weakness in internal control. We consider two
types of growth. The first, ACQUISITION VALUE, is the aggregate dollar value of
acquisitions that result in at least 50 percent ownership of the acquired company in years t
and t–1 scaled by the acquiring firm’s year t market capitalization, as reported by the SDC
Platinum database. Our second rapid growth measure, EXTREME SALES GROWTH, is
an indicator variable that is equal to one if the firm’s year-over-year sales growth is in the
highest quintile of sales growth for their industry, and zero otherwise.16 We expect rapid
growth to be positively associated with internal control weaknesses.
Similarly, we expect firms undergoing restructuring to have relatively more internal
control weaknesses. First, restructuring often results in the downsizing of departments,
the loss of experienced employees, and general disarray during and after the re-engineering
of the firm—the internal control system must be updated to match the new organiza-
tional structure. Second, restructuring typically involves many difficult accrual
estimations and adjustments (e.g., impairment of goodwill; see also Dechow and Ge,
2006). Insufficient staff and more accounting estimation likely lead to more internal
control deficiencies. For example, Nortel Networks disclosed a lack of compliance
with the established procedures for monitoring and adjusting balances relating to
restructuring charges as a material weakness in their June 2003 10-Q. As larger
restructurings are likely to cause proportionally larger challenges for internal control,
we define RESTRUCTURING CHARGE as a continuous variable, equal to aggregate
restructuring charges in years t and t–1 scaled by the firm’s year t market capitalization.17
We expect material weaknesses to be more prevalent as firms record larger restructuring
charges.
Finally, we expect corporate governance to play a role in a firm’s internal control
quality. Krishnan (2005) finds that firms with more effective audit committees report fewer
internal control problems in their 8-Ks when reporting an auditor change. We expect a
well-governed firm to exhibit fewer material weaknesses, all else equal.18 We measure
corporate governance with the measure developed by Brown and Caylor (2006).
GOVERNANCE SCORE is a composite measure of 51 factors encompassing eight
corporate governance categories: audit, board of directors, charter/bylaws, director
education, executive and director compensation, ownership, progressive practices, and

16
Results are not sensitive to the calculation of this metric. We also consider sales growth as a continuous
variable, unadjusted sales growth (as both a continuous variable and an indicator variable), and each sales growth
metric from year t–2 to year t. We opt to use the extreme growth indicator variable to focus on those firms that
have ‘‘abnormal’’ sales growth that is most likely to cause firms to outgrow their internal controls. We use the one-
year time horizon to minimize the data requirement.
17
Results hold if we use current year’s restructuring charge or three years’ aggregate restructuring charge. We
also consider indicator variables for non-zero restructuring charges, and charges of at least one percent and five
percent of market value. Restructuring charges remain a positive predictor of material weaknesses if an indicator
variable is used, though results weaken slightly, suggesting that the likelihood of a weakness is increasing in the
magnitude of the restructuring charge.
18
In this paper, we use material weakness disclosures to proxy for the existence of an underlying internal control
problem. Governance may also affect the choice to disclose a known deficiency (e.g., Ashbaugh-Skaife et al.,
2007), thus biasing against our stated hypothesis. Krishnan (2005) examines audit committee effectiveness.
However, her sample predates the new requirements pertaining to audit committee independence and required
financial expertise (see Klein, 2003, for an overview of these requirements). Thus, there is sufficient variation in her
data for a powerful test of her hypothesis. However, in recent years, the variation in audit committee effectiveness
has declined. For example, the mean percentage of independent audit board members is 90 percent for 2002.
ARTICLE IN PRESS
J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223 203

Table 1
Variable definitions and expected relation with the probability of disclosing a material weakness

Variable Predicted Calculation


direction

MARKETCAP  The log of share price x number of shares outstanding [data item
#25  data item #199]
FIRM AGE  The log of the number of years the firm has CRSP data

AGGREGATE LOSS + An indicator variable equal to one if earnings before


extraordinary items [data item #18] in years t and t–1 sum to less
than zero, and zero otherwise
BANKRUPTCY RISK + The decile rank of the percentage probability of bankruptcy from
the default hazard model prediction based on Shumway (2001)
SPEs + The log of the number of special purpose entities associated with
the firm in year t
SEGMENTS + The log of the sum of the number of operating and geographic
segments reported by the Compustat Segments database for the
firm in year t

FOREIGN + An indicator variable equal to one if the firm has a non-zero


TRANSACTIONS foreign currency translation [data item #150] in year t, and zero
otherwise
ACQUISITION VALUE + The aggregate dollar value of acquisitions that result in at least 50
percent ownership of the acquired company in years t and t–1
scaled by the acquiring firm’s year t market capitalization
EXTREME SALES + An indicator variable that is equal to one if year-over-year
GROWTH industry-adjusted sales growth [data item #12] falls into the top
quintile, and zero otherwise

RESTRUCTURING + The aggregate restructuring charges [data item #376  1] in


CHARGE years t and t–1 scaled by the firm’s year t market capitalization
GOVERNANCE SCORE  A composite measure of 51 factors encompassing eight corporate
governance categories: audit, board of directors, charter/bylaws,
director education, executive and director compensation,
ownership, progressive practices, and state of incorporation
developed by Brown and Caylor (2006)

state of incorporation.19 We summarize each of our directional predictions and variable


measurements in Table 1.

3. Data, sample selection, and material weakness classifications

3.1. Data and sample selection

As mentioned above, material weaknesses in internal control have only been widely
disclosed in SEC filings since August of 2002. Since November 2003, Compliance Week
19
We thank Larry Brown and Marcus Caylor for providing the governance data.
ARTICLE IN PRESS
204 J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223

(a website dedicated to Sarbanes-Oxley related compliance issues) has been collecting


and publishing monthly reports on firms that disclose internal control deficiencies. We
aggregate these monthly Compliance Week disclosures, obtaining 877 individual
disclosures. We include only those firms that classify their internal control problem(s) as
a material weakness, the most severe internal control deficiency, eliminating 239 non-
material weakness disclosures.20 In the event of a subsequent disclosure by the same
firm, we retain the disclosure to aid in our type analysis, but do not include the firm
multiple times in our sample, descriptive statistics, or statistical tests. There are 97 such
duplicate disclosures.21 To supplement our sample, we also search 10-Ks in the EDGAR
database from August 2002 (the enactment date of Section 302) to August 2005 using the
keyword ‘‘material weakness’’ and identify an additional 429 material weakness firms,
thereby identifying a total of 970 distinct firms that disclosed at least one material
weakness from August 2002 to 2005.22 We then obtain financial data from the 2003
annual Compustat database and eliminate 191 firms that have insufficient Compustat data,
resulting in a final sample of 779 material weakness firms.23 In 175 instances, the material
weakness firm does not have available data for 2003, but does have data available
on Compustat for a prior year. In these cases, we take the most recent data available (2002
for 163 firms, 2001 for nine firms, and 2000 for three firms). In summary, our sample
period is much longer than that of other concurrent work as we examine three full years,
including both Section 302 and 404 disclosures. For example, Ashbaugh-Skaife et al.
(2007) limit their analysis to Section 302 disclosures in the year prior to the effective date of
Section 404.
For our control firms, we use all 2003 Compustat firms with available market value of
equity and earnings before extraordinary items that are not in our material weakness
sample and did not appear on Compliance Week as having reported a less severe
significant deficiency. We use the entire non-material weakness population as our control
group, rather than a matched sample, to avoid choice-based sample bias, which can lead to
biased parameters and probability estimates (Palepu, 1986). We summarize our sample
collection procedure in Table 2. It is likely that most material weaknesses were disclosed, as
it is a criminal offense for managers to conclude that controls are effective when they
have knowledge of a material weakness. However, the use of the proxy of a disclosure
of a material weakness versus the true underlying existence of a weakness is a limitation of
our study.

20
We read through each SEC filing to ensure a material weakness in internal control is disclosed. We eliminate
non-material weakness disclosures, or significant deficiencies, in order to focus on the most severe types of internal
control problems. Compliance Week discontinued tracking significant deficiencies that did not reach the level of a
material weakness in March 2005. Therefore, no conclusions can be drawn regarding the proportion of material
weaknesses to all significant deficiencies.
21
Duplicates include those instances where a parent and subsidiary both file with the SEC and report the same
material weakness. In these cases we include only the parent company; we found no instances where a parent and
subsidiary reported different material weaknesses.
22
Note that Compliance Week only tracks firms belonging to the Russell 3000 beginning in January 2005; only
using internal control deficiency disclosures collected by Compliance Week would introduce a sample bias.
Therefore, we supplement our data by searching the EDGAR database.
23
Sixty-five companies in our final sample disclosed a material weakness related to lease accounting in early
2005, which primarily results from the views expressed by the Office of the Chief Accountant of the SEC in a
February 7, 2005 letter to the AICPA. Our results are almost identical if we exclude these companies from our
sample.
ARTICLE IN PRESS
J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223 205

Table 2
Sample selection procedure

Test sample

Total disclosures on Compliance Week from November 2003 to August 2005 877
Less: Internal control disclosures that were not material weaknesses (239)
Duplicate disclosures (97)
Total distinct material weakness disclosures on Compliance Week 541
Add: Additional material weakness disclosures via a 10-K search from August 2002 to 2005 429
Total identified material weakness disclosures 970
Less: Firms without Compustat data (191)

Final sample of distinct firms that disclosed a material weakness from August 2002 to 2005 779
Control Sample
Total firms on 2003 Compustat (with non-missing market value of equity and earnings before 5,935
extraordinary items)
Less: Firms with identified material weakness disclosures (779)
Firms with identified internal control disclosures that were not material weaknesses and (109)
subsidiaries of firms in our test sample that also disclosed an internal control deficiency
Final sample of control firms 5,047

3.2. Material weakness classifications

Although we focus on material weaknesses, the most severe internal control problems,
these disclosures vary widely with respect to both the severity and underlying
reason. Therefore, we partition our sample based on the description of each material
weakness found in the SEC filing. Butler et al. (2004) illustrate the importance
of examining subsets of common disclosures (in their case qualified audit opinions).
In their setting, they find that a subset of going concern opinions appears to drive
the association between abnormal accruals and qualified audit opinions. Similarly, we wish
to explore how determinants differ among the various types of material weakness
disclosures.
We have two classification schemes, one based on the severity of the internal control
problem and the second based on the stated reason for the internal control problem.
We then examine whether the underlying determinants of internal control problems
vary based on severity or reason for the weakness. To determine whether a material
weakness is severe, we follow the logic put forth by Moody’s, the bond rating
company. Moody’s proposes that material weaknesses fall into one of two categories.
Account-specific material weaknesses relate to controls over specific account balances
or transaction-level processes. Moody’s suggests that these types of material weaknesses
are identifiable by auditors through substantive testing and thus do not represent as
serious a concern regarding the reliability of the financial statements. Company-level
material weaknesses, however, relate to more macro-level controls such as the control
environment or the overall financial reporting process, which auditors may not be able
ARTICLE IN PRESS
206 J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223

to effectively ‘‘audit around.’’ Moody’s suggests that company-level material weaknesses


call into question not only management’s ability to prepare accurate financial reports but
also its ability to control the business (Doss and Jonas, 2004). We detail this classification
scheme further in Appendix B.24
Though we do not have explicit hypotheses and emphasize that this work is
exploratory in nature, it seems reasonable to believe that less severe, account-specific
problems would be more common in larger, more mature firms that are dealing with
complex accounting issues, often arising from rapidly changing operations. In contrast,
more severe, company-wide material weaknesses would seem to be more prevalent in
smaller, younger, financially weaker, and poorly governed firms that do not have the
resources (or the proper board oversight) to maintain a comprehensive system of internal
control.
Our second classification scheme is based on the company’s stated reason for the
material weakness. For example, some material weaknesses are related to staffing issues,
while others pertain to overseas operations or inter company transactions. For this
classification scheme, we create three categories of material weaknesses: Staffing,
Complexity, and General. These categories are based on the main types of problems
that we noted in our collection of material weaknesses. For example, there are a large
number of material weaknesses associated with segregation of duties, and clearly we
anticipate that these weaknesses will have fundamentally different determinants than those
related to revenue recognition policies in overseas subsidiaries.
The common staffing issues are ‘‘inadequate segregation of duties,’’ ‘‘inadequate
qualified staffing and resources,’’ or ‘‘lack of a full-time CFO.’’ Complexity issues include,
for example, trouble interpreting and applying complex accounting standards, such
as those related to hedging and derivatives. General issues are usually related to
deficient revenue recognition policies or control weaknesses in the period-end reporting
process. This final type of weakness allows for greater opportunity to manage earnings.
Note that in the first classification scheme (severity), we classify each material weakness
company as either account-specific or company-level (i.e., the categories are mutually
exclusive). In the second classification scheme, each company can have multiple
deficiency types. For example, one company can have both a staff shortage deficiency
and a material weakness related to applying complex accounting standards of hedge
transactions. Again, examples of our material weakness classification schemes are included
in Appendix B.
As mentioned, it seems likely that the determinants will differ across these three
categories. Staffing issues (including segregation of duties problems) will probably be much
more likely for smaller, younger, and financially weaker firms that lack resources to hire or
train appropriate personnel. Firms with complexity-related weaknesses will likely be
larger, older firms with highly diversified and complex operations. Finally, since the
General category includes firms with revenue recognition problems, we expect that these
firms may have weaker corporate governance.

24
In some cases, it is straightforward to categorize a company as having company-level material weaknesses; for
example, when ‘‘ineffective control environment’’ or ‘‘management override’’ is specifically identified as a material
weakness in the disclosure—consider, for example, Hollinger, Inc., displayed in Appendix A. However, most
disclosures are not so forthcoming. Thus, if a firm has a material weakness related to at least three account-specific
problems, we classify the firm as having a company-level material weakness.
ARTICLE IN PRESS
J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223 207

4. Results

4.1. Univariate analysis and descriptive statistics

Table 3 presents descriptive statistics on the characteristics of material weakness


firms and control firms. It also presents one-tailed tests of differences between the two
groups using both t-tests and Wilcoxon rank-sum tests. For ease of interpretation, each
summary statistic for the four logged variables (MARKETCAP, FIRM AGE, SPEs, and
SEGMENTS) is converted to an unlogged amount in Table 3. For example, the mean of
the log of MARKETCAP of 5.193 is unlogged to generate the 180.082 value that is
presented in the first column.
First, the results on firm size, as measured by the log of market capitalization, seem to be
mixed. The mean value is weakly significantly smaller for material weakness firms (using a
t-test), but the median value is actually slightly larger for the control firms, and the
Wilcoxon rank-sum test indicates an insignificant difference between the two groups.
Although these univariate results are mixed, we hypothesize that firm size is a determinant
of internal control quality after controlling for complexity. Therefore, we re-examine this
variable in our multivariate analysis below.
Next, as predicted, firms with material weaknesses appear to be younger than
the other Compustat firms. The mean material weakness firm has been publicly
traded for 8.3 years, while the mean control firm has been traded for 9.1 years.
Our financial health measures, AGGREGATE LOSS and BANKRUPTCY RISK,
both indicate that firms disclosing material weaknesses in internal control are
significantly weaker financially than the average Compustat firm, as expected. Our
three complexity measures, the log of the number of sponsored special purpose
entities (SPEs), the log of the total number of operating and geographic segments
(SEGMENTS), and the existence of a foreign currency adjustment (FOREIGN
TRANSACTIONS), are all higher for material weakness firms, providing preliminary
support for our hypothesis that accounting complexity creates internal control
challenges.
There is also preliminary evidence that rapid growth is a determinant of internal
control problems. ACQUISITION VALUE is significantly higher for material
weakness firms than our control firms (0.033 versus 0.023), and material weakness
firms are also more often in the highest quintile of industry-adjusted sales growth
(0.222 versus 0.196). RESTRUCTURING CHARGE for material weakness firms
is more than double that for the control firms, lending support for the idea that
significant restructuring, with its disruptions to established processes, personnel
turnover, and difficult accounting estimates, leads to internal control problems.
Finally, there is some support that well-governed firms have fewer material weak-
nesses, with the t-test and Wilcoxon rank-sum test both generating weakly significant
results.
In general our univariate results are consistent with our hypotheses outlined in Section
2.2. However, as evidenced in Table 4, many of our measures are correlated with one
another. For example, MARKETCAP, FIRM AGE, SPEs, and SEGMENTS are all
positively correlated with one another, and each is negatively correlated with
AGGREGATE LOSS and BANKRUPTCY RISK. We examine these potential determi-
nants further using a multivariate analysis below.
208
J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223
Table 3
Descriptive statistics of material weakness firms versus 2003 Compustat firms

Material weakness firms 2003 Compustat firms (excluding material weakness firms)

Variable Mean Median Std. Dev. 25% 75% Predicted Mean Median Std. Dev. 25% 75%
difference

ARTICLE IN PRESS
MARKETCAP ($M) 180.082 224.420 7.862 49.242 660.317 o 204.196c 214.022 12.584 36.795 1,186.259
FIRM AGE 8.274 8.000 2.557 5.000 17.000 o 9.105a 10.000a 2.620 5.000 19.000
AGGREGATE LOSS 0.497 0.000 0.500 0.000 1.000 4 0.373a 0.000a 0.484 0.000 1.000
BANKRUPTCY RISK 4.875 5.000 2.945 2.000 7.000 4 4.437a 4.000a 2.855 2.000 7.000
SPEs 1.879 0.000 2.784 0.000 3.000 4 1.690a 0.000a 2.682 0.000 2.000
SEGMENTS 3.440 4.000 2.048 2.000 6.000 4 2.752a 3.000a 2.213 0.000 5.000
FOREIGN TRANSACTIONS 0.208 0.000 0.406 0.000 0.000 4 0.156a 0.000a 0.363 0.000 0.000
ACQUISITION VALUE 0.033 0.000 0.105 0.000 0.000 4 0.023a 0.000c 0.081 0.000 0.000
SALES GROWTH 0.222 0.000 0.416 0.000 0.000 4 0.196b 0.000b 0.397 0.000 0.000
RESTRUCTURING CHARGE 0.032 0.000 0.106 0.000 0.008 4 0.013a 0.000a 0.064 0.000 0.000
GOVERNANCE SCORE 22.325 22.000 3.563 20.000 25.000 o 22.667c 22.000c 3.476 20.000 25.000

All variables are described in Table 1. The t-test of means uses the pooled method when the underlying variances are equal and the Satterthwaite method when they
are unequal. There are a maximum of 779 material weakness firm observations and 5,047 control firm observations. Each of the continuous variables is winsorized at
1% and 99% to mitigate outliers. For ease of interpretation, each summary statistic for the four logged variables (MARKETCAP, FIRM AGE, SPEs, and
SEGMENTS) is converted to an unlogged amount when presented above.
a,b, or c
Significantly different from Material Weakness group at a one-tailed p-value p0.01, 0.05, or 0.10, respectively, under a t-test (shown on mean value above) or
Wilcoxon rank-sum test (shown on median value above).
Table 4
Pearson correlation matrix

J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223


MARKETCAP FIRM AGG. BANKR. SPEs SEG- FRGN. ACQ. EXTR. RESTR. GOV.
AGE LOSS RISK MENTS TRANS. VALUE SLS GRO CHG SCORE

MW 0.017 0.035 0.087 0.054 0.036 0.096 0.048 0.039 0.022 0.090 0.035
(0.1872) (0.0134) (0.0001) (0.0004) (0.0054) (0.0001) (0.0002) (0.0028) (0.0969) (0.0001) (0.1164)
MARKETCAP 0.103 0.402 0.505 0.351 0.379 0.204 0.103 0.008 0.106 0.253

ARTICLE IN PRESS
(0.0001) (0.0001) (0.0001) (0.0001) (0.0001) (0.0001) (0.0001) (0.5386) (0.0001) (0.0001)
FIRM AGE 0.125 0.052 0.156 0.217 0.044 0.025 0.171 0.020 0.281
(0.0001) (0.0005) (0.0001) (0.0001) (0.0016) (0.0768) (0.0001) (0.1473) (0.0001)
AGGREGATE LOSS 0.160 0.175 0.048 0.002 0.044 0.047 0.195 0.132
(0.0001) (0.0001) (0.0003) (0.8665) (0.0008) (0.0005) (0.0001) (0.0001)
BANKRUPTCY RISK 0.092 0.252 0.141 0.025 0.162 0.146 0.085
(0.0001) (0.0001) (0.0001) (0.0955) (0.0001) (0.0001) (0.0002)
SPEs 0.152 0.017 0.112 0.066 0.013 0.088
(0.0001) (0.1893) (0.0001) (0.0001) (0.3275) (0.0001)
SEGMENTS 0.335 0.084 0.016 0.102 0.146
(0.0001) (0.0001) (0.2394) (0.0001) (0.0001)
FOREIGN TRANSACTIONS 0.013 0.002 0.052 0.055
(0.3205) (0.8848) (0.0001) (0.0130)
ACQUISITION VALUE 0.148 0.004 0.039
(0.0001) (0.7493) (0.0811)
EXTREME SALES GROWTH 0.078 0.017
(0.0001) (0.4570)
RESTRUCTURING CHARGE 0.054
(0.0162)

There are a maximum of 5,826 observations. All variables are described in Table 1. Each of the continuous variables is winsorized at 1% and 99% to mitigate outliers.

209
ARTICLE IN PRESS
210 J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223

4.2. Multivariate analysis

We model the probability of disclosing a material weakness in internal control over


financial reporting as a function of the above-mentioned firm characteristics using a
logistic regression with the following constructs:

ProbðMW Þ ¼ f ðb0 þ b1 SIZE þ b2 FIRM AGE þ b3 FINANCIAL HEALTH


þ b4 COMPLEXITY þ b5 RAPID GROWTH
þ b6 RESTRUCTURING þ b7 GOVERNANCE
þ SK
k¼1 gk INDUSTRY Þ. ð1Þ

MW is an indicator variable that is equal to one if the firm disclosed a material weakness
in internal control, and zero if the firm is a control firm. Though not tabulated, 16 industry
indicator variables are also included to capture the tendency of material weakness firms to
cluster by industry (Ge and McVay, 2005).25 We present three regression specifications
since two of our variables (BANKRUPTCY RISK and GOVERNANCE SCORE) have
limited availability.
Referring to the first column of results in Table 5, all of the coefficients are in the
predicted direction and statistically significant at p-values less than 0.05 under one-tailed
tests. Overall, the joint marginal effect of the model (i.e., the change in the predicted
probability of a material weakness when altering the independent variables in the predicted
direction between the 1st and 3rd quartiles or between zero and one for indicator variables)
more than quintuples the predicted probability of a material weakness. The predicted
probability rises from 4.17 percent to 23.66 percent (not tabulated), showing that the
model’s predictive ability is economically significant.
The second column of results incorporates the financial health variable, BANK-
RUPTCY RISK. This causes a loss of about 13 percent of our sample. Results are similar
in this specification, though ACQUISITION VALUE weakens (X2 statistic of 2.18). As
predicted, a higher probability of bankruptcy is positively associated with reporting a
material weakness (X2 statistic of 3.81). The joint marginal effect of this specification raises
the predicted probability of a material weakness from 3.75 percent to 26.41 percent (not
tabulated).
Our final column of results includes GOVERNANCE SCORE. Note that the number of
available observations drops dramatically, falling about 58 percent from the second
column specification. Not surprisingly, results are somewhat weaker in this specification.
MARKETCAP, SPEs, and ACQUISITION VALUE are no longer significant, likely due
to both the lower power and possible size bias introduced by requiring corporate
governance data. These private datasets tend to provide data for the largest firms, which
reduce the variation in size and size-related variables. GOVERNANCE SCORE is not

25
Industry classifications are compiled using the following SIC codes: Agriculture 100–999; Mining: 1000–1299,
1400–1999; Food: 2000–2199; Textiles: 2200–2799; Drugs: 2830–2839, 3840–3851; Chemicals: 2800–2829,
2840–2899; Refining: 1300–1399, 2900–2999; Rubber: 3000–3499; Industrial: 3500–3569, 3580–3659; Electrical:
3660–3669, 3680–3699; Miscellaneous Equipment: 3700–3839, 3852–3999; Computers: 3570–3579, 3670–3679,
7370–7379; Transportation: 4000–4899; Utilities: 4900–4999; Retail: 5000–5999; Banks: 6000–6999; Services:
7000–7369, 7380–8999; Miscellaneous: 9000–9999. We also estimate the regression without the industry indicator
variables, finding almost identical results.
ARTICLE IN PRESS
J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223 211

Table 5
Logistic regression of the probability of disclosing a material weakness

Dependent variable ¼ MW

Independent variables Predicted sign Logit estimate Logit estimate Logit estimate
(X2) (X2) (X2)

INTERCEPT 2.182*** 2.200*** 2.650***


(129.95) (57.61) (18.35)
MARKETCAP  0.080*** 0.105*** 0.018
(12.20) (11.78) (0.11)
FIRM AGE  0.136** 0.134** 0.338***
(8.30) (6.12) (11.76)
AGGREGATE LOSSES + 0.438*** 0.331** 0.329*
(19.95) (9.48) (3.76)
BANKRUPTCY RISK + 0.038* 0.090**
(3.81) (7.89)
SPEs + 0.161*** 0.153*** 0.040
(13.70) (11.38) (0.35)
SEGMENTS + 0.269*** 0.303*** 0.291**
(14.04) (15.21) (5.83)
FOREIGN TRANSACTIONS + 0.311** 0.320** 0.448**
(7.60) (7.31) (6.10)
ACQUISITION VALUE + 0.763* 0.682 0.185
(3.07) (2.18) (0.07)
EXTREME SALES GROWTH + 0.227* 0.262* 0.311*
(4.45) (5.06) (2.83)
RESTRUCTURING CHARGE + 1.184** 2.148** 2.510**
(6.62) (10.89) (5.73)
GOVERNANCE SCORE  0.011
(0.26)

Industry indicator variables Included Included Included


Number of material weakness obs 707 627 273
Number of total observations 4,984 4,333 1,841
Likelihood ratio w2 (p-value) 230.247 229.690 126.085
(0.001) (0.001) (0.001)

MW is an indicator variable that is equal to one if the firm disclosed a material weakness in internal control from
August 2002 to August 2005, and zero otherwise. All other variables are defined in Table 1. Each of the
continuous variables is winsorized at 1% and 99% to mitigate outliers.
***, **, * PrXX2 of 0.001, 0.01, 0.05, respectively.

statistically significant in this specification (X2 statistic of 0.26). We continue to consider


this variable when examining specific types of material weaknesses below.
Overall, both our univariate and multivariate findings support our hypotheses outlined
in Section 2.2. We find that material weaknesses in internal control are more likely for
firms that are smaller, younger, financially weaker, more complex, growing rapidly,
and/or undergoing restructuring. These findings are consistent with firms experiencing
challenges with their financial reporting controls in the face of a lack of resources, complex
accounting issues, and/or a rapidly changing business environment. We do not find a
significant relation between material weakness disclosures and corporate governance;
ARTICLE IN PRESS
212 J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223

Table 6a
Descriptive statistics of material weakness firms by severity of deficiency

Account-specific Company-level
(maximum of 491 (maximum of 286
observations) observations)

Variable Mean Median Predicted Mean Median


difference

MARKETCAP ($M) 208.819 264.407 4 142.836a 179.408a


FIRM AGE 8.617 8.000 4 7.695c 8.000c
AGGREGATE LOSS 0.420 0.000 o 0.627a 1.000a
BANKRUPTCY RISK 4.770 5.000 o 5.079 5.000c
SPEs 1.941 0.000 4 1.786 0.000c
SEGMENTS 3.458 4.000 4 3.438 4.000
FOREIGN TRANSACTIONS 0.226 0.000 4 0.178c 0.000c
ACQUISITION VALUE 0.035 0.000 4 0.029 0.000
EXTREME SALES GROWTH 0.226 0.000 4 0.213 0.000
RESTRUCTURING CHARGE 0.031 0.000 ? 0.034 0.000
GOVERNANCE SCORE 22.284 22.000 4 22.408 22.000

There are a total of 779 material weakness firms, of which two had insufficient information to classify. All
variables are described in Table 1. The t-tests of means use the pooled method when the underlying variances are
equal and the Satterthwaite method when they are unequal. The difference tests for classification scheme #1
compare the variables for account-specific material weaknesses to the variables for company-level weaknesses;
these classifications are mutually exclusive. Each of the continuous independent variables is winsorized at 1% and
99% to mitigate outliers. For ease of interpretation, each summary statistic for the four logged variables
(MARKETCAP, FIRM AGE, SPEs, and SEGMENTS) is converted to an unlogged amount when presented
above.
a, b, or c
Significantly different from account-specific group at a one-tailed p-value p0.01, 0.05, or 0.10,
respectively, under a t-test (shown on mean value above) or Wilcoxon rank-sum test (shown on median value
above).

however, this may be a result of low power due to sample size limitations. In sum, not only
are the variables generally statistically significant in the directions predicted, the economic
significance of the overall models is quite high, increasing the predicted probability of a
material weakness from about 4 percent to about 26 percent.

4.3. Material weakness type analysis

The above analysis combines all material weakness disclosure types. However, as
previously mentioned, material weaknesses vary widely both in severity and the underlying
reason for the weakness. Some occur in auditable accounts (account-specific), which can be
easily corrected through adjusting entries, while others may be more pervasive and difficult
to mitigate by additional auditor testing (company-level). Some are associated with staffing
concerns (Staffing), and others with complex accounting issues (Complexity). Thus we
explore how the determinants differ by material weakness type (outlined in Appendix B
and discussed in Section 3.2).
We begin by providing descriptive statistics by material weakness type in Tables 6a and
6b and comparing the variables among the groups using t-tests and Wilcoxon rank-sum
ARTICLE IN PRESS
J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223 213

Table 6b
Descriptive statistics of material weakness firms by reason for deficiency

Staffing Complexity General


(maximum of 251 (maximum of 347 (maximum of 519
observations) observations) observations)

Variable Mean Median Mean Median Mean Median

MARKETCAP ($M) 103.516c1,g1 140.724c1,g1 235.316s1,g5 263.304s1,g5 181.877s1,c5 224.420s1,c5


FIRM AGE 7.308c1 8.000c5 8.754s1,g5 8.000s5,g5 7.877c5 8.000c5
c1,g10 c1,g10
AGGREGATE LOSS 0.582 1.000 0.467s1,g5 0.000s1,g5 0.516s10,c5
1.000s10,c5
BANKRUPTCY RISK 5.163 6.000 4.943 5.000 4.879 5.000
SPEs 1.477c1,g1 0.000c1,g1 2.121s1,g5 0.000s1,g5 1.841s1,c5 0.000s1,c5
SEGMENTS 3.249c1,g5 4.000c1,g1 3.691s1 4.000s1 3.573s5 4.000s1
c1 c1
FOREIGN 0.195 0.000 0.265s1,g5 0.000s1,g5 0.210c5
0.000c5
TRANSACTIONS
ACQUISITION VALUE 0.032 0.000c10 0.035 0.000s10,g5 0.031 0.000c5
EXTREME SALES 0.271c5,g5 0.000c1,g5 0.200s5 0.000s1 0.223s5 0.000s5
GROWTH
RESTRUCTURING 0.025c5,g10 0.000c10 0.036s5 0.000s10 0.036s10 0.000
CHARGE
GOVERNANCE SCORE 22.304 22.000 22.329 22.000 22.106 22.000

All variables are described in Table 1. The t-tests of means use the pooled method when the underlying variances
are equal and the Satterthwaite method when they are unequal. The difference tests for classification scheme #2
compare the variables for one group of material weakness firms (e.g., Staffing) to another group (e.g.,
Complexity). As this classification scheme is not mutually exclusive, if a firm is classified into two groups for
scheme #2, it is dropped from the test calculation between those two groups. This causes a drop in the sample sizes
actually tested and results in testing values that are somewhat different from those shown above. Each of the
continuous independent variables is winsorized at 1% and 99% to mitigate outliers. For ease of interpretation,
each summary statistic for the four logged variables (MARKETCAP, FIRM AGE, SPEs, and SEGMENTS) is
converted to an unlogged amount when presented above.
c1,5, or 10
Significantly different from Complexity group at a one-tailed p-value p0.01, 0.05, or 0.10, respectively,
under a t-test (shown on mean value above) or Wilcoxon rank-sum test (shown on median value above).
g1,5, or 10
Significantly different from General group at a one-tailed p-value p0.01, 0.05, or 0.10, respectively,
under a t-test (shown on mean value above) or Wilcoxon rank-sum test (shown on median value above).
s1,5, or 10
Significantly different from Staffing group at a one-tailed p-value p0.01, 0.05, or 0.10, respectively, under
a t-test (shown on mean value above) or Wilcoxon rank-sum test (shown on median value above).

tests of the differences. As in Table 3, each summary statistic for the four logged variables
(MARKETCAP, FIRM AGE, SPEs, and SEGMENTS) is converted to an unlogged
amount in Tables 6a and 6b for ease of interpretation. For the first classification scheme,
presented in Table 6a there are 491 account-specific material weakness firms, 286 firms in
the company-level group, and two firms with insufficient information for classification
(recall that these two groups are mutually exclusive). When contrasting within the material
weakness sample, firms reporting more serious company-level weaknesses tend to be
smaller, younger, and financially weaker (larger AGGREGATE LOSS and BANK-
RUPTCY RISK). Firms reporting company-level weaknesses also appear to be less
complex (smaller SPEs, SEGMENTS, and FOREIGN TRANSACTIONS) and growing
less rapidly (smaller ACQUISITION VALUE and EXTREME SALES GROWTH). The
differences in restructuring charges and governance quality between the groups are not
statistically significant under either the t-test or Wilcoxon rank-sum test. In general, firms
ARTICLE IN PRESS
214 J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223

with company-wide problems seem to lack the resources or experience to maintain


comprehensive control systems.
In contrast, the firms with transaction-level weaknesses tend to be more mature (i.e.,
they are larger, older, and financially stronger), have more diversified and complex
operations, and have higher internal and acquisition-related growth. The complexity of
these firms’ operating environments, along with the rapid change evidenced by merger and
acquisition activity and high sales growth, seems to hinder them in maintaining adequate
account-specific internal controls.
Table 6b presents our second classification scheme, where 251 of the disclosures
relate to staffing (e.g., segregation of duties), 347 to complexity (e.g., consolidating foreign
subsidiaries), and 519 were more general in nature (e.g., insufficient reconciliation
procedures). Since these groups are not mutually exclusive, if a firm is classified into two
groups it is dropped from the t-test of means and Wilcoxon rank-sum calculation between
those two groups. We compare the means and ranks of our variables for one group (e.g.,
Staffing) to another group (e.g., Complexity).
Not surprisingly, when comparing Staffing problems to Complexity and General
weaknesses, we find that Staffing problems are far more pervasive in smaller and
younger firms. These firms also tend to be the weakest financially, with the highest
incidence of losses and the highest average probability of bankruptcy. Firms with
staffing problems also tend to have the least complex organizations (lowest SPEs,
SEGMENTS, and FOREIGN TRANSACTIONS), which is perhaps related to their
smaller size. Firms with staffing issues have the highest sales growth and incur the
lowest average restructuring charges, which indicates that these firms are in the early
stages of the business life cycle. These findings suggest that, when a firm is smaller,
younger, financially weaker, and in a higher growth stage, it might not be able to commit
sufficient resources for qualified staff and training. Weak internal controls follow as a
direct consequence.
Conversely, firms with disclosures related to Complexity are the largest and oldest firms,
have the lowest incidence of losses, and have the most diversified operations (highest SPEs,
SEGMENTS, and FOREIGN TRANSACTIONS). They also have the highest level of
acquisitions (ACQUISITION VALUE), the lowest extreme sales growth, and the highest
governance quality (GOVERNANCE SCORE), perhaps reflecting their larger size and
more mature status.
The final group, General, includes firms with a wide range of problems ranging from
lack of documentation for transactions, deficiencies in the closing process, problems
with revenue recognition, and transaction-specific issues, to problems with the overall
tone set by management. As expected, with such a wide range of issues included in this
group, few of the variables stand out from the other two groups. Most interestingly, given
the fact that these firms have internal control problems more related to earnings
management-type activities (e.g., revenue recognition issues or outright fraud), is that the
GOVERNANCE SCORE is the lowest of the three groups (although not statistically
significantly lower). Perhaps poor governance is contributing to these types of internal
control problems.
In Table 7, we re-estimate the logistic regression Eq. (1) with the alternate dependent
variables, account-specific and company-level. In Table 8, our alternate dependent
variables are Staffing, Complexity, and General. Note that the regression now compares
the firms in each group to the original Compustat control group and not to the other
ARTICLE IN PRESS
J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223 215

Table 7
Logistic regression of the probability of disclosing a material weakness by severity of deficiency

Dependent variable ¼ Dependent variable ¼


MW_ACCOUNT-SPECIFIC MW_COMPANY-LEVEL

Independent variables Logit estimate Logit estimate


(X2) (X2)

INTERCEPT 2.579*** 3.434***


(54.54) (53.29)
MARKETCAP 0.116*** 0.081
(10.21) (2.66)
FIRM AGE 0.099 0.211**
(2.38) (5.67)
AGGREGATE LOSS 0.093 0.777***
(0.52) (20.02)
BANKRUPTCY RISK 0.032 0.054*
(1.88) (3.06)
SPEs 0.151** 0.172*
(7.85) (5.29)
SEGMENTS 0.261** 0.393***
(7.90) (9.59)
FOREIGN TRANSACTIONS 0.451*** 0.102
(10.49) (0.27)
ACQUISITION VALUE 0.874* 0.263
(2.72) (0.107)
EXTREME SALES GROWTH 0.337** 0.111
(6.00) (0.34)
RESTRUCTURING CHARGE 2.627*** 1.475
(12.48) (2.40)
Industry indicator variables Included Included
Number of material weakness obs 415 212
Number of total observations 4,121 3,918
Likelihood ratio w2 (p-value) 160.997 123.721
(0.001) (0.001)

MW_DESCRIPTOR is an indicator variable that is equal to one if the firm disclosed a material weakness (either
Account-specific or Company-level) in internal control from August 2002 to 2005, and zero if the firm did not
disclose a material weakness. Following Moody’s, we define account-specific material weaknesses as those that are
at the transaction level, and company-level material weaknesses as those that are more entity-wide control
problems. All other variables are defined in Table 1. Each of the continuous variables is winsorized at 1% and
99% to mitigate outliers.
***, **, * PrXw2 of 0.001, 0.01, 0.05, respectively.

groups, as in Tables 6a and 6b. In both tables, we exclude GOVERNANCE SCORE as this
variable severely limits the sample size. In results not tabulated, GOVERNANCE SCORE
is insignificant in all specifications except when predicting control weaknesses over revenue
recognition, which we discuss below.
The first column of results in Table 7 has transaction-level problems as the dependent
variable, where MW_ACCOUNT-SPECIFIC is an indicator variable that is equal to
one if the firm disclosed a material weakness related to a specific account, and zero if
the firm did not disclose a material weakness. MARKETCAP, SPEs, SEGMENTS,
ARTICLE IN PRESS
216 J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223

Table 8
Logistic regression of the probability of disclosing a material weakness by reason for deficiency

Dependent variable ¼ Dependent variable ¼ Dependent variable ¼


MW_STAFFING MW_COMPLEXITY MW_GENERAL

Independent variables Logit estimate Logit estimate Logit estimate


(X2) (X2) (X2)

INTERCEPT 3.419*** 3.333*** 2.545***


(44.90) (65.15) (54.21)
MARKETCAP 0.101* 0.097* 0.118***
(3.57) (5.25) (10.36)
FIRM AGE 0.240** 0.054 0.183**
(6.41) (0.49) (8.08)
AGGREGATE LOSSES 0.397* 0.356** 0.353**
(4.52) (5.64) (7.71)
BANKRUPTCY RISK 0.078** 0.056* 0.034
(5.46) (4.01) (2.15)
SPEs 0.006 0.218*** 0.148**
(0.00) (13.34) (7.17)
SEGMENTS 0.211 0.302** 0.407***
(2.39) (7.65) (19.00)
FOREIGN TRANSACTIONS 0.353* 0.692*** 0.233*
(3.08) (19.46) (2.74)
ACQUISITION VALUE 0.986 0.777 0.248
(1.55) (1.48) (0.19)
EXTREME SALES 0.321* 0.240 0.265*
GROWTH (2.69) (2.12) (3.70)
RESTRUCTURING 0.970 2.089** 2.447***
CHARGE (0.72) (6.26) (12.54)
Industry indicator variables Included Included Included
Number of MW observations 180 295 417
Number of total observations 3,886 4,001 4,123
Likelihood ratio w2 (p-value) 113.169 151.629 186.538
(0.001) (0.001) (0.001)

MW_DESCRIPTOR is an indicator variable that is equal to one if the firm disclosed a material weakness
described as due to staffing, business complexity, or that was more general in nature (STAFFING,
COMPLEXITY and GENERAL, respectively), and zero if the firm did not disclose a material weakness. All
other variables are defined in Table 1.
***, **, * PrXX2 of 0.001, 0.01, 0.05, respectively.

FOREIGN TRANSACTIONS, ACQUISITION VALUE, EXTREME SALES GROWTH


and RESTRUCTURING CHARGE are all significant in the hypothesized directions at
p-values less than 0.05 under one-tailed tests in this estimation. Companies with account-
specific control problems do not appear to be financially weaker than the average
Compustat firm (AGGREGATE LOSS is insignificant, while BANKRUPTCY RISK is
fairly weak; one-tailed p-value ¼ 0.085, not tabulated). The high level of complexity and
diversification for these firms (higher SPEs, SEGMENTS, and FOREIGN TRANSAC-
TIONS) and rapidly changing environment (higher ACQUISITION VALUE, EXTREME
SALES GROWTH and RESTRUCTURING CHARGE) may be contributing to more
ARTICLE IN PRESS
J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223 217

complex transactions (e.g., consolidation issues, restructuring-related accruals, etc.),


resulting in account-specific material weaknesses.
Company-level material weaknesses, presented in the final column in Table 7, tend to be
slightly smaller (one-tailed p-value ¼ 0.052, not tabulated), younger, and financially
weaker than the average Compustat firm. Interestingly, two of our three complexity
measures load, SPEs and SEGMENTS, suggesting that decentralization might exacerbate
serious control issues. Finally, a rapidly changing environment does not appear to lead to
serious company-level control problems (ACQUISITION VALUE and EXTREME
SALES GROWTH are insignificant, while RESTRUCTURING CHARGE is only weakly
significant [one-tailed p-value ¼ 0.061, not tabulated]).
In sum, firms with company-level material weaknesses are in some respects quite
different from those that report account-specific weaknesses. Firm age and financial health
are stronger predictors of more pervasive, company-level material weaknesses, while
foreign transactions and rapid growth appear to be stronger predictors of account-specific
problems. Decentralized operations appear to contribute to both types of problems, albeit
for different reasons. This characteristic introduces both accounting complexity (account-
specific) and a potential lack of oversight (company-level).
Table 8 presents our second classification scheme, based on whether the weakness was
categorized as Staffing, Complexity, and/or General. The first column of results in Table 8
has Staffing as the dependent variable, where MW_STAFFING is an indicator variable
that is equal to one if the firm disclosed a material weakness related to staffing issues, and
zero if the firm did not disclose a material weakness.
First, both MARKETCAP and FIRM AGE are negative predictors of staffing
problems, consistent with the notion that small and young firms are less likely to have
sufficient resources or experience to develop high-quality accounting controls. These
firms also likely have smaller accounting departments, exacerbating segregation
of duties issues. Interestingly, of our three complexity measures, only FOREIGN
TRANSACTIONS is significant at a p-value p0.05, perhaps reflecting the additional
staffing challenges introduced by having international operations. While rapid
growth appears to be a weak predictor of staffing problems, restructurings are not,
which is inconsistent with the notion that restructurings create staffing issues by laying
off experienced workers. It appears that firms undergoing restructuring are large
enough to avoid these issues. Overall, these results suggest that smaller, younger, and
financially weaker firms may lack the resources to maintain adequate staffing levels
and/or training, especially when the firm is rapidly growing and/or has multinational
operations.
Turning to the second column of results, where the dependent variable is Complexity,
firm age is no longer statistically significant, in contrast to the results for Staffing
firms. Firms with complexity-related problems are smaller, financially weaker, more
operationally and geographically diverse, and more likely to be undergoing restructuring
than the average Compustat firm. These results are fairly intuitive. Complexity issues are
driven by heterogeneous operations (SPEs, SEGMENTS, and FOREIGN TRANSAC-
TIONS), while firms with fewer resources (MARKETCAP, AGGREGATE LOSSES,
and BANKRUPTCY RISK) are least able to implement effective controls. Finally,
complexity-related weaknesses appear to be strongly associated with restructuring
charges. As previously mentioned, restructuring charges result in many complicated
estimations (e.g., Dechow and Ge, 2006) and also may lead to more complicated tax
ARTICLE IN PRESS
218 J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223

Table 9
Logistic regression of the probability of disclosing a material weakness related to revenue recognition problems

Dependent variable ¼
MW_REVENUE_RECOGNITION

Independent variables Logit estimate Logit estimate


(X2) (X2)

INTERCEPT 4.450*** 3.372***


(43.87) (53.29)
MARKETCAP 0.091 0.092
(1.80) (0.67)
FIRM AGE 0.380*** 0.383*
(10.33) (2.96)
AGGREGATE LOSS 0.143 0.190
(0.38) (0.27)
BANKRUPTCY RISK 0.050 0.070
(1.50) (1.13)
SPEs 0.205* 0.048
(3.61) (0.07)
SEGMENTS 0.720*** 0.624*
(15.47) (4.93)
FOREIGN TRANSACTIONS 0.374 0.689*
(2.58) (4.04)
ACQUISITION VALUE 2.294 6.025*
(2.29) (2.95)
EXTREME SALES GROWTH 0.523* 0.761*
(4.90) (4.41)
RESTRUCTURING CHARGE 3.418*** 2.072
(9.76) (0.96)
GOVERNANCE SCORE 0.093*
(3.59)
Industry indicator variables Included Included
Number of material weakness obs 113 49
Number of total observations 3,819 1,617
Likelihood ratio w2 (p-value) 121.002 72.093
(0.001) (0.001)

MW_REVENUE_RECOGNITION is an indicator variable that is equal to one if the firm disclosed a material
weakness described as due to revenue recognition problems, and zero if the firm did not disclose a material
weakness. All other variables are defined in Table 1.
***, **, * PrXX2 of 0.001, 0.01, 0.05, respectively.

issues (113 of our complexity-related firms mentioned tax issues). In general, firms with
complexity-related problems seem to be smaller, financially troubled firms that have highly
diverse operations.
Finally, we examine General material weaknesses in our final column of results.
Each of the determinants examined is statistically significant with a one-tailed p-value
p0.05, with the exception of BANKRUPTCY RISK (one-tailed p-value ¼ 0.071, not
tabulated) and ACQUISITION VALUE (one-tailed p-value ¼ 0.333, not tabulated).
These general findings are not surprising as the ‘‘General’’ classification contains many
different issues.
ARTICLE IN PRESS
J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223 219

We further decompose General firms into those that are related to revenue recognition
problems, which comprise 131 of the 417 General firms. As Table 9 shows, we find that firm
size and financial health are no longer strong predictors. The financial health result suggests
that some firms might choose to have weak internal control over revenue recognition in
order to manage earnings. We investigate this further by adding in GOVERNANCE
SCORE, which to this point has been an insignificant predictor in all multivariate analyses.
We find that GOVERNANCE SCORE is a negative and significant predictor of material
weaknesses related to revenue recognition (see second column of results).
Overall, the determinants of material weaknesses appear to be dependent on the types
of material weaknesses disclosed. Firms with transaction-level weaknesses seem to
have more diversified and complex operations and are undergoing significant changes.
Firms with more serious, company-level problems tend to be younger and weaker
financially. Consistent with each firm facing their own set of internal control challenges,
staffing-related internal control problems are more pervasive in small, financially weak
firms, while firms attributing their internal control weaknesses to complexity-related issues
are more likely to have diversified and complex operations and to be undergoing rapid
change.

4.4. Restatements

Auditing Standard No. 2 issued by the PCAOB indicates that a restatement of


previously issued financial statements is a ‘‘strong indicator that a material weakness in
internal control over financial reporting exists.’’ The purpose of this paper is to
examine the determinants of the underlying internal control problem itself, and not any
resulting restatement. Thus, as a sensitivity analysis, we eliminate those firms that
announced a restatement in connection with the material weakness disclosure or
within one year of the material weakness disclosure (hereafter contemporaneous
restatement). To identify contemporaneous restatement firms, we search Lexis-Nexis
Academic Universe using the keyword ‘‘restate!’’ for each of our material weakness
firms. To ensure our identification procedure is as accurate as possible, we look up each
firm on EGDAR to ensure the company’s name has not changed, and search on all
possible names in the event of a name change. We also search through each of the material
weakness disclosures to ensure that any firms announcing a restatement in that disclosure
were also in our restatement sample. We identify 393 firms with contemporaneous
restatements (50.4 percent of our 779 material weakness firms). We then exclude these
restatement firms, and re-estimate our main regression (Table 5, Column 2). The results
(not tabulated) for these non-restatement firms are similar to those reported—each
coefficient is in the same predicted direction and all one-tailed p-values remain at levels of
0.06 or less.

5. Summary and conclusions

The recent passage of the Sarbanes-Oxley Act in 2002 marks the first time that all SEC
registrants must publicly disclose material weaknesses in internal control over financial
reporting. Past research on internal control has been limited to deficiency disclosures from
firms that changed their auditors (this was the only prior public disclosure necessary for all
SEC registrants), which created a very limited source of information (e.g., Krishnan, 2005).
ARTICLE IN PRESS
220 J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223

Using a more comprehensive sample of mandatory material weakness disclosures made


pursuant to Sections 302 and 404 of Sarbanes-Oxley from August 2002 to August 2005, we
examine the determinants of material weaknesses. We show that material weaknesses in
internal control are more likely for firms that are smaller, less profitable, more complex,
growing rapidly, or undergoing restructuring. These findings are consistent with firms
struggling with their financial reporting controls in the face of a lack of resources, complex
accounting issues, and a rapidly changing business environment. We also document that
these determinants vary in strength depending on the type of material weakness disclosed.
This finding is informative for future research examining the reactions to, and implications
of, internal control deficiencies.
A potential limitation of this study is the short time frame over which we gather our data
and conduct our tests. It is hard to determine how the material weakness disclosures of the
first few years of the new Sarbanes-Oxley regime will compare to future periods when both
management and auditors are more familiar with the process of implementing, evaluating,
and reporting on internal control. In addition, although we attempt to be comprehensive in
collecting material weakness disclosures from August 2002 to 2005, it is still possible that
some firms did not discover or disclose their material weaknesses, thus causing us to under-
identify our true sample. Depending on the systematic characteristics of such firms, this
could have an effect on our results. For example, it is possible that the materiality
thresholds for determining whether or not a firm has a material weakness in internal
control vary among the sample firms. Future research might attempt to model this
materiality decision (Messier et al., 2005).

Appendix A. Examples of material weakness disclosures

America West Airlines (AWA) (12/31/2004 10-K)

Management concluded that AWA’s fuel-hedging transactions did not qualify for hedge
accounting under US generally accepted accounting principles and that the Company’s
financial statements for prior periods required restatement to reflect the fair value of fuel-
hedging contracts in the balance sheets and statements of stockholders equity and
comprehensive income for Holdings and AWA. These accounting errors were the result of
deficiencies in its internal control over financial reporting from the lack of effective reviews
of hedge transaction documentation and of quarterly mark-to-market accounting entries
on open fuel hedging contracts by personnel at an appropriate level.

Comstock Homebuilding Company, Inc. (8/13/2004 S-1)

In connection with their audits of our financial statements, our independent auditors
have reported certain conditions, which together constitute a material weakness in the
internal controls over our ability to produce timely and accurate financial statements y
The conditions resulting in the material weakness gave rise to a number of adjustments
under generally accepted accounting principles, and adjustments relating to the
completeness and accuracy of certain underlying data, which materially changed our
financial statements. Our independent auditors also identified a need to add to the staff
and strengthen the overall skills of our accounting department.
ARTICLE IN PRESS
J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223 221

Hollinger International, Inc. (12/31/2003 10-K)

The Company’s management concluded that the following material weaknesses in the
Company’s internal controls and ineffectiveness in the design and operation of the
Company’s disclosure controls and procedures, among others, existed during the year
ended or as of 31 December 2003:

 The ‘‘tone from the top’’ established by the former executive officers was inappropriate
to the establishment of an environment in which strong systems of internal controls and
disclosure controls and procedures are encouraged.
 Certain former executive officers of the Company, who were also executive officers at
the Company’s various controlling stockholders, did not participate in open and timely
communication with those responsible for the preparation of corporate reports or with
the Board of Directors, in particular its independent members.
 The management and corporate organizational structures facilitated extraction of assets
from the Company by way of related party transactions to benefit direct and indirect
controlling stockholders. y

The above pervasive weaknesses directly or indirectly led to other material weaknesses
or significant deficiencies in internal controls, such as inadequate documentation of
business processes and internal controls.

Appendix B. Examples of material weakness classification schemes

First classification scheme

Account-specific or transaction-level material weaknesses

(1) Inadequate internal controls for accounting for loss contingencies, including bad debts
(2) Deficiencies in the documentation of a receivables securitization program
(3) No adequate internal controls over the application of new accounting principles or the
application of existing accounting principles to new transactions

Company-level material weaknesses

(1) Override by senior management


(2) Ineffective control environment

Second classification scheme

Staffing:

(1) Inadequate qualified staffing and resources leading to the untimely identification and
resolution of certain accounting and disclosure matters and failure to perform timely
and effective reviews
(2) The need to increase the training of the financial staff
(3) Weak internal controls and procedures relating to separation of duties
ARTICLE IN PRESS
222 J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223

Complexity:

(1) Inconsistencies in the application of company policies among business units and
segments
(2) Material weaknesses in the interpretation and application of complex accounting
standards, such as standards related to hedge transactions

General:

(1) Weak internal controls related to contracting practices


(2) Deficiencies related to the design of policies and execution of processes related to
accounting for transactions
(3) Deficiencies in the period-end reporting process

References

Ashbaugh-Skaife, H., Collins, D., Kinney, W., 2007. The discovery and reporting of internal control deficiencies
prior to SOX-mandated audits. Journal of Accounting and Economics, forthcoming.
Beneish, D., Billings, M., Hodder, L., 2006. International control weaknesses and information uncertainty.
Working paper, Indiana University.
Brown, L., Caylor, M., 2006. Corporate governance and firm performance. Journal of Accounting and Public
Policy 25, 409–434.
Bryan, S., Lilien, S., 2005. Characteristics of firms with material weaknesses in internal control: an assessment of
Section 404 of Sarbanes Oxley. Working paper, Wake Forest University and Baruch College.
Bushman, R., Chen, Q., Engle, E., Smith, A., 2004. Financial accounting information, organizational complexity
and corporate governance systems. Journal of Accounting and Economics 37, 167–201.
Butler, M., Leone, A., Willenborg, M., 2004. An empirical analysis of auditor reporting and its association with
abnormal accruals. Journal of Accounting and Economics 37, 139–165.
Chan, K., Farrell, B., Lee, P., 2005. Earnings management and return-earnings association of firms reporting
material internal control weaknesses under Section 404 of the Sarbanes-Oxley Act. Working paper, Pace
University.
Dechow, P., Ge, W., 2006. The persistence of earnings and cash flows and the role of special items: implications
for the accrual anomaly. Review of Accounting Studies 11, 253–296.
DeFond, M., Jiambalvo, J., 1991. Incidence and circumstances of accounting errors. The Accounting Review 66,
643–655.
DeFond, M., Raghunandan, K., Subramanyam, K.R., 2002. Do non-audit service fees impair auditor
independence? Evidence from going concern audit opinions. Journal of Accounting Research 40,
1247–1274.
Doss, M., Jonas, G., 2004. Section 404 reports on internal control: impact on ratings will depend on nature of
material weaknesses reported. Moody’s Investors Service, Global Credit Research. October.
Doyle, J., Ge, W., McVay, S., 2007. Accruals quality and internal control over financial reporting. The
Accounting Review, forthcoming.
Frankel, R., Johnson, M., Nelson, K., 2002. The relation between auditors’ fees for non-audit services and
earnings management. The Accounting Review 77, 71–103.
Ge, W., McVay, S., 2005. The disclosure of material weaknesses in internal control after the Sarbanes-Oxley Act.
Accounting Horizons 19, 137–158.
Geiger, M., Taylor, P., 2003. CEO and CFO certifications of financial information. Accounting Horizons 17,
357–368.
Glass Lewis and Company Yellow Card Internal Control Compliance Report, September 23, 2004.
Hogan, C., Wilkins, M., 2005. Internal control weaknesses and earnings management. Working paper, Southern
Methodist University and Texas A&M University.
ARTICLE IN PRESS
J. Doyle et al. / Journal of Accounting and Economics 44 (2007) 193–223 223

Kinney, W., 2000. Research opportunities in internal control quality and quality assurance. Auditing 19, 83–90.
Kinney, W., 2001. Accounting scholarship: what is uniquely ours? The Accounting Review 76, 275–284.
Kinney, W., Maher, M., Wright, D., 1990. Assertions-based standards for integrated internal control. Accounting
Horizons 4, 1–8.
Kinney, W., McDaniel, L., 1989. Characteristics of firms correcting previously reported quarterly earnings.
Journal of Accounting and Economics 11, 71–93.
Klein, A., 2003. Likely effects of stock exchange governance proposals and Sarbanes-Oxley on corporate boards.
Accounting Horizons 17, 343–355.
Krishnan, J., 2005. Audit committee quality and internal control: an empirical analysis. The Accounting Review
80, 649–675.
Mayper, A., 1982. Consensus of auditors’ materiality judgments of internal accounting control weaknesses.
Journal of Accounting Research 20, 773–783.
Mayper, M., Doucet, S., Warren, C., 1989. Auditors’ materiality judgments of internal accounting control
weaknesses. Auditing: A Journal of Practice and Theory 9, 72–86.
McMullen, D., Raghunandan, K., Rama, D.V., 1996. Internal control reports and financial reporting problems.
Accounting Horizons 10, 67–75.
Messier, W., Martinov-Bennie, N., Eilifsen, A., 2005. A review and integration of empirical research on
materiality: two decades later. Auditing: A Journal of Practice and Theory 24, 153–187.
Ogneva, M., Raghunandan, K., Subramanyam, K.R., 2006. Internal control weakness and cost of equity:
evidence from SOX Section 404 disclosures. Working paper, University of Southern California and Florida
International University.
Palepu, K., 1986. Predicting takeover targets. Journal of Accounting and Economics 8, 3–35.
Public Company Accounting Oversight Board (PCAOB), 2004. Auditing Standard No. 2—An Audit of Internal
Control over Financial Reporting Performed in Conjunction with an Audit of Financial Statements.
Richardson, S., Tuna, I., Wu, M., 2003. Predicting earnings management: the case of earnings restatements.
Working paper, University of Pennsylvania and Hong Kong University of Science and Technology.
SEC, 1988. Disclosure Amendments to Regulation S-K, form 8-K and Schedule 14A Regarding Changes in
Accountants and Potential Opinion Shopping Situations, Financial Reporting Release No. 31 (April), SEC
Docket 1140–1147. Washington, DC.
SEC, 2002. Certification of Disclosure in Companies’ Quarterly and Annual Reports, Final Rule 33-8124 (August
29), Washington, DC.
SEC, 2003. Management’s Reports on Internal Control Over Financial Reporting and Certification of Disclosure
in Exchange Act Periodic Reports, Final Rule 33-8238 (June 5), Washington, DC.
SEC, 2004. Management’s Reports on Internal Control Over Financial Reporting and Certification of Disclosure
in Exchange Act Periodic Reports—Frequently Asked Questions, (October 6), Washington, DC.
Shumway, T., 2001. Forecasting bankruptcy more accurately: a simple hazard model. The Journal of Business 74,
101–124.
Stice, J., 1991. Using financial and market information to identify pre-engagement factors associated with lawsuits
against auditors. The Accounting Review 66, 516–533.